The Dynamical Cluster Approximation (DCA) is modified to include disorder. The DCA incorporates non-local corrections to local approximations such as the Coherent Potential Approximation (CPA) by mapping the lattice problem with disorder, and in the thermodynamic limit, to a self-consistently embedded finite-sized cluster problem. It satisfies all of the characteristics of a successful cluster approximation. It is causal, preserves the point-group and translational symmetry of the original lattice, recovers the CPA when the cluster size equals one, and becomes exact as $N_c\to\infty$. We use the DCA to study the Anderson model with binary diagonal disorder. It restores sharp features and band tailing in the density of states which reflect correlations in the local environment of each site. While the DCA does not describe the localization transition, it does describe precursor effects of localization.